Surface roughness modifications (cont'd) #2045

slevis-lmwg · 2023-06-28T22:57:10Z

Description of changes

This PR continues from where #1596 left off.

Specific notes

Contributors other than yourself, if any:
@ekluzek @RonnyMeier

CTSM Issues Fixed (include github issue #):
Fixes #1596
Fixes #1316

Are answers expected to change (and if so in what way)?
Yes

Any User Interface Changes (namelist or namelist defaults changes)?
Yes, NLCOMP diffs in the test-suites

Testing performed, if any:
Test-suite results reported below

…ime output (the latter is still in development). This version of the code is running stable.

…lowing meeting at January 12 2022.

… Meier

…ORTRAN code

…2 code

…ghness as htop isn't set yet. Also use a different threshold for snomelt_accum suggested by Ronny Meier

… of a negative number being done, in this case set wc to zero and continue

…ss is used to prevent a numerical error that kills the run in calc of WetBulb temp because of negative relative humidity

…he year

…SCOMP#1374

Resolved conflicts: doc/ChangeLog doc/ChangeSum

ekluzek

OK, awesome. I have some questions and things to do, more parameters mostly. Some of this could be made into an issue and done later.

We'll decide on this when we talk it over together.

ekluzek · 2023-08-17T20:40:49Z

src/biogeophys/FrictionVelocityMod.F90

+       select case (z0param_method)
+       case ('ZengWang2007')
+          if (frac_sno(c) > 0._r8) then
+             if(use_z0m_snowmelt) then


There was some discussion on this in #1596. The question I wonder about is this as presented adds the "use_z0m_snowmelt option for ZengWang2007 (and use_z0m_snowmelt is a new option for this new work). So I wonder if we want to enable this option for the old structure? I kind of think we should only let it work on the new Meir2022 option.

The discussion in question is here: #1596 (comment)

For code maintenance unless this option is needed for science we'd like to simplify it and not add this option for ZengWang.

@olyson or @swensosc do you have opinions on how useful this would be for running with ZengWang?

If we simplify though we should change the buildnml so that you can NOT use this option with ZengWang.

I'm not aware of any need for this. Might be better to simplify.

ekluzek · 2023-08-17T20:44:38Z

src/biogeophys/FrictionVelocityMod.F90

+                if ( snomelt_accum(c) < 1.e-5_r8 )then
+                    z0mg(c) = exp(-b1_param * rpi * 0.5_r8 + b4_param) * 1.e-3_r8
+                else
+                    z0mg(c) = exp(b1_param * (atan((log10(snomelt_accum(c))+0.23_r8)/0.08_r8)) + b4_param) * 1.e-3_r8


The 0.23_r8 is repeated 4 times here and should be made into a parameter either in the code or params file.

Also the 0.08

ekluzek · 2023-08-17T20:47:13Z

src/biogeophys/CanopyFluxesMod.F90

+
+         case ('Meier2022')
+            lt = max(0.00001_r8,elai(p)+esai(p))
+            displa(p) = htop(p) * (1._r8 - (1._r8 - exp(-(cd1_param * lt)**0.5_r8)) / (cd1_param*lt)**0.5_r8)


I wonder if these **0.5 terms should be changed to sqrt? Although that will likely change answers. So maybe we make it an issue and do it as an answer changing tag?

Let's not do this. But, possibly add some comments about it....

Addressed with comments

ekluzek · 2023-08-17T20:48:46Z

src/biogeophys/CanopyFluxesMod.F90

+                      *z0v_c(patch%itype(p)) * lt * 0.25_r8
+            U_ustar = U_ustar_ini
+
+            do while (delt > 0.0001_r8)


There's this 0.0001 here and the 0.00001 above that I wonder if they should be parameters? Although they don't seem to be repeated....

Since, they aren't repeated we think we'll leave them alone.

This expression is confusing though. For example the **(-0.5) which means 1/sqrt(...).

We think adding a comment about this would be helpful.

Hmm, I didn't realize that they were different, and made them both 1.e-4_r8. I should probably back this change out...

ekluzek · 2023-08-17T20:53:55Z

bld/namelist_files/namelist_defaults_ctsm.xml

@@ -214,6 +214,8 @@ attributes from the config_cache.xml file (with keys converted to upper-case).
 <zetamaxstable phys="clm4_5"                                   >2.0d00</zetamaxstable>
 <zetamaxstable phys="clm5_0"                                   >0.5d00</zetamaxstable>
 <zetamaxstable phys="clm5_1"                                   >0.5d00</zetamaxstable>


I think we should remove this line, as the user should have to choose between ZengWang and Meier. I'm not sure it matters, but I think it's clearer with it removed...

We should make sure z0param_method HAS to be selected in the buildnml.

ekluzek · 2023-08-17T20:56:01Z

src/biogeophys/BareGroundFluxesMod.F90

+            case ('Meier2022')
+
+               ! After Yang et al. (2008)
+               z0hg_patch(p) = 70._r8 * nu_param / ustar(p) * exp( -beta_param * ustar(p)**(0.5_r8) * (abs(tstar))**(0.25_r8))


The 70 is repeated and should become a parameter. Actually I think this expression is repeated and maybe should become a function. This come later...

Also the **0.5...

We think we should leave the **0.5 and **0.25 should stay as they are. The **0.25 should probably stay that way as that's better than sqrt(sqrt(...)), so the other one should as well.

ekluzek · 2023-08-17T20:58:08Z

src/biogeophys/BareGroundFluxesMod.F90

+               forc_hgt_u_patch(p) = forc_hgt_u_patch(g) + z0mg_patch(p) + displa(p)
+               forc_hgt_t_patch(p) = forc_hgt_t_patch(g) + z0hg_patch(p) + displa(p)
+               forc_hgt_q_patch(p) = forc_hgt_q_patch(g) + z0qg_patch(p) + displa(p)
+            end if
            thvstar = tstar*(1._r8+0.61_r8*forc_q(c)) + 0.61_r8*forc_th(c)*qstar


This 0.61 is somewhat outside the scope here, but should become a parameter. We should maybe make an issue for that.

ekluzek · 2023-08-17T21:00:24Z

src/biogeophys/HumanIndexMod.F90

@@ -1014,6 +1014,15 @@ subroutine Wet_BulbS (Tc_6,rh,wbt)
 ! !LOCAL VARIABLES:
 !EOP
 !
+#ifndef NDEBUG


Do we really want this #ifdef here?

This was on purpose. But, I'm not sure if really should be done.

Was it expensive?

This was on purpose. But, I'm not sure if really should be done (the #if...)

Was it expensive? No not more expensive than the expression below which is expensive, and can blow up for bad values.

And this WetBulb stuff is already known to be expensive, hence we don't turn it on that often.

So final bottom-line remove the #ifndef and the #endif

ekluzek · 2023-08-17T22:51:47Z

src/biogeophys/BareGroundFluxesMod.F90

+
+               ! After Yang et al. (2008)
+               z0hg_patch(p) = 70._r8 * nu_param / ustar(p) * exp( -beta_param * ustar(p)**(0.5_r8) * (abs(tstar))**(0.25_r8))
+


Let's remove this commented out code. We can't maintain it when commented out. And without scientific effort we can't add it, as we don't know if it really offers scientific value.

ekluzek · 2023-08-17T22:53:10Z

src/biogeophys/CanopyFluxesMod.F90

@@ -1007,7 +1066,12 @@ subroutine CanopyFluxes(bounds,  num_exposedvegp, filter_exposedvegp,
            ! changed by K.Sakaguchi from here
            ! transfer coefficient over bare soil is changed to a local variable
            ! just for readability of the code (from line 680)
-            csoilb = vkc / (params_inst%a_coef * (z0mg(c) * uaf(p) / 1.5e-5_r8)**params_inst%a_exp)
+            ! RM: Does this need to be updated if Ya08 is used too? Proposed formulation (definitely double-check!)


Remove this commented out alternative, that we can't maintain.

ekluzek · 2023-08-17T22:54:33Z

src/main/clm_varctl.F90

+  !  Surface roughness parameterization
+  !----------------------------------------------------------
+
+  character(len=64), public :: z0param_method


This should have a comment...

ekluzek · 2023-08-17T22:56:43Z

This has more work to be done on it, so won't be done immediately. But, we'll schedule it for next week....

ekluzek · 2023-08-17T22:58:28Z

When I do a diff I see a lot of red for white-space at the ends of lines. It would be good to remove that for the new code coming in. The red makes me hate that white-space at the end.

slevis-lmwg · 2023-08-18T01:28:21Z

Addressed most suggestions and now running two preliminary tests before submitting test-suites:
PASS ./create_test ERS_Ly3.f10_f10_mg37.I2000Clm51BgcCrop.cheyenne_intel.clm-Meier2022_surf_rough
PASS ./create_test ERP_P72x2_Lm25.f10_f10_mg37.I2000Clm51BgcCrop.cheyenne_intel.clm-monthly -c /glade/p/cgd/tss/ctsm_baselines/ctsm5.1.dev134

slevis-lmwg · 2023-08-21T21:36:09Z

Test-suites comparing against dev134: OK

expected failures
FSURDATMODIFYCTSM_D_Mmpi-serial_Ld1.5x5_amazon.I2000Clm50SpRs.cheyenne_intel (expected conda fails to load for SystemTests #2111)
NLCOMP diffs
field lists differ

Resolved conflicts: doc/ChangeLog doc/ChangeSum

apcraig · 2023-08-24T22:14:21Z

I'd like to ask a question about this PR.

I am adding CTSM to RASM (CESM1.2). I'm currently doing so with release-clm5.0.35, which I think is the last release from several months ago. Anyway, we need to couple the (logz0) to WRF. Several years I did the same with an earlier version of CTSM (for the NNA project, still ongoing), but I'm trying to shift to a more recent and robust release. In the older version, there was a subroutine called SetActualRoughnessLengths which computed z0's from a call in clm_drv. In release-clm5.0.35, that subroutine is gone and it looks like z0's are only computed in Fates, if Fates is on (just from looking at the code). Is that correct?

In our test run, use_fates=.false. and it looks like z0's are never computed. Several questions.

Does this PR and recent changes address this shortcoming? Are z0's now computed when fates is off?
Should I be running with fates on?
More generally, would you recommend another version more recent than release-clm5.0.35 for us to use? I prefer to use a more robust version than some recent hash in master, but there are trade-offs.

Thanks.

slevis-lmwg · 2023-08-24T23:27:53Z

Hi @apcraig ,

My interpretation of this PR's mods:
This PR modified how z0 terms were calculated (eg see description of changes in the ChangeLog ). So z0 terms were already being calculated a different way before this PR (eg see changes in BaregroundFluxes, BiogeophysPreFluxCalcs and elsewhere). Looking at this PR's mods may help you more easily find where z0 is calculated. This PR does not turn on the new z0 calculations by default, so it keeps answers unchanged from the last baseline.

You should have access to z0 calculations with or without Fates. I do not know why Fates does its own calculation of z0. Maybe it was easier (or better) than using the clm calculation.

As you suggest, it may not make sense to work with this most recent version UNLESS you wish to use the new z0 calculations (z0param_method = "Meier2022").

apcraig · 2023-08-25T02:15:15Z

Thanks @slevis-lmwg. OK, I'll have a look at this version. I think an important point is that in release-clm5.0.35, z0 is only calculated by fates. At some point since that version, it sounds like this was updated and z0 is calculated with or without fates. It sounds like I don't need the latest version, but I need some version since release-clm5.0.35. Can you recommend a particular ctsm tags or hash that is relatively robust and tested that might meet my needs, I assume something newer than release-clm5.0.35 and something older than the current master. Thanks!

slevis-lmwg · 2023-08-25T02:46:16Z

When I type git grep z0 in a copy of clm5.0.00 that I have, I see calculations of z0, so I wonder why release-clm5.0.35 is different. I don't feel I have followed the model development closely enough to be able to recommend the right version for your work.

Ronny Meier and others added 30 commits March 25, 2021 08:37

Adapted model to machine and added soil column separation and local t…

7e5d438

…ime output (the latter is still in development). This version of the code is running stable.

Saving current state

db0b16f

Saving state

492c158

Saving state

def8ee3

Saving state

4018d56

Saving state

4a52aa7

Saving state

e78b02f

Saving state

37d7d0c

Saving state

d218ff5

State at first submission

fc20d00

For PR

2e7b2c3

Removing statements of sensitivity experiments. Add more comments fol…

864f336

…lowing meeting at January 12 2022.

Small correction to proposed statement in CanopyFluxes

8318cad

Add user-mod directories for the surface roughness changes from Ronny…

c611d6e

… Meier

Remove use_z0v_forest because it's not connected to anything in the F…

9ae6e30

…ORTRAN code

Handle very small values of htop and snowmelt_accum with new Meier202…

0ee8c69

…2 code

Don't set z0mg and displa on first step of run for RonnyMeier2022 rou…

02be2d8

…ghness as htop isn't set yet. Also use a different threshold for snomelt_accum suggested by Ronny Meier

Skip first few steps as htop isn't set until after first do_alb step

0b7fd39

Correct subscript

0e71cb2

Add write of rh if negative or over 100 when DEBUG on

84396e6

Write out warning if ustar*thvstar is positive to prevent a cube root…

23601a7

… of a negative number being done, in this case set wc to zero and continue

Set human stress indices to calculate NONE when Meier surface roughne…

4eabfbb

…ss is used to prevent a numerical error that kills the run in calc of WetBulb temp because of negative relative humidity

Fix subscript

f644624

Rename test mods to highlight these are non-crop cases

32dbda8

Add tests with Meier surface roughness

b9d402a

Add a bit to not die if htop is zero and it's crop and beginning of t…

8789ca2

…he year

Correct new testname, and add two tests to the expected fails

821df3c

Change MeierXXXX namelist to Meier2022

d3dee39

Remove commented out line

95fb900

Remove local time history changes as they are treated seperately in E…

b0f2856

…SCOMP#1374

Merge remote-tracking branch 'escomp/master' into meier_main_pr1596

53722d2

Resolved conflicts: doc/ChangeLog doc/ChangeSum

slevis-lmwg mentioned this pull request Aug 17, 2023

Surface roughness modifications #1596

Closed

This comment was marked as outdated.

Sign in to view

ekluzek requested changes Aug 17, 2023

View reviewed changes

ekluzek reviewed Aug 17, 2023

View reviewed changes

wwieder mentioned this pull request Aug 18, 2023

Documentation of new features needed #2113

Open

12 tasks

Revisions part 1 in response to Erik's review

1d962dd

slevis-lmwg added 6 commits August 21, 2023 15:41

Revisions part 2: add missing comment

59745d0

Revert delt_threshold param back to hardwired values

fea92f2

Ensure that ZengWang2007 and use_z0m_snowmelt NOT true simultaneously

7a39c65

Merge remote-tracking branch 'escomp/master' into meier_main_pr1596

9e54cd5

Resolved conflicts: doc/ChangeLog doc/ChangeSum

Removed white space that appears red in git diff

0875c51

Merge remote-tracking branch 'escomp/master' into meier_main_pr1596

f206914

Resolved conflicts: doc/ChangeLog doc/ChangeSum

slevis-lmwg merged commit 0e7457f into ESCOMP:master Aug 23, 2023
2 checks passed

slevis-lmwg deleted the meier_main_pr1596 branch August 23, 2023 17:59

slevis-lmwg mentioned this pull request Oct 25, 2023

Meier roughness changes answers on processor count #2219

Closed

samsrabin added enhancement new capability or improved behavior of existing capability science Enhancement to or bug impacting science labels Aug 8, 2024


		! After Yang et al. (2008)
		z0hg_patch(p) = 70._r8 * nu_param / ustar(p) * exp( -beta_param * ustar(p)*(0.5_r8) (abs(tstar))**(0.25_r8))

Surface roughness modifications (cont'd) #2045

Surface roughness modifications (cont'd) #2045

Conversation

slevis-lmwg commented Jun 28, 2023 • edited Loading

Description of changes

Specific notes

This comment was marked as outdated.

ekluzek left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ekluzek commented Aug 17, 2023

ekluzek commented Aug 17, 2023

slevis-lmwg commented Aug 18, 2023 • edited Loading

slevis-lmwg commented Aug 21, 2023 • edited Loading

apcraig commented Aug 24, 2023

slevis-lmwg commented Aug 24, 2023

apcraig commented Aug 25, 2023

slevis-lmwg commented Aug 25, 2023

slevis-lmwg commented Jun 28, 2023 •

edited

Loading

slevis-lmwg commented Aug 18, 2023 •

edited

Loading

slevis-lmwg commented Aug 21, 2023 •

edited

Loading